Perineural invasion (PNI) is an important pathologic feature of cervical cancer that is associated with poor prognosis and provides key information for clinical decisions. A better understanding of the molecular mechanisms underlying PNI could lead to improved patient treatment strategies. Here, we generated whole-exome, whole-genome, and RNA sequencing data from tumors and matched normal clinical samples of 45 patients with cervical cancer and performed a comparative analysis between 23 PNI and 22 non-PNI tumors. A robust machine learning approach identified a three-gene expression signature of MT1G, NPAS1, and SPRY1 that could predict the tumor PNI status with high accuracy, which was validated using an independent cohort (18 PNI and 19 non-PNI). Loss-of-function FBXW7 mutations were identified as driver events for PNI that lead to increased MYC activity and an immunosuppressive tumor microenvironment. Finally, a deep learning model for predicting drug efficacy over patients' transcriptomic data revealed OTX015, a BET inhibitor, as a promising treatment that targets mutated FBXW7 PNI tumors. This study provides a rich resource for elucidating the molecular mechanisms of PNI tumors, laying a critical foundation for developing effective diagnostic and therapeutic strategies for PNI tumors in cervical cancer.
Significance: Generation of a rich resource for characterizing the molecular basis of perineural invasion in tumors lays a critical foundation for developing effective diagnostic and therapeutic strategies in cervical cancer. This article is part of a special series: Driving Cancer Discoveries with Computational Research, Data Science, and Machine Learning/AI .
©2025 American Association for Cancer Research.